06-05 · Calf feeding and growth
CSV · Dictionary · Provenance · Inspection starter · Analysis plan
Scientific context and your question
This is a hypothetical teaching study inspired by Jensen, Jensen & Vestergaard (2020). All supplied observations are synthetic; they are not the paper’s raw data or a reconstruction of its results. Source reading provides context and a design comparison, not an analysis to copy.
Explore a question about growth under the two milk-feeding strategies. Choose the outcome summary, follow-up time or trajectory feature and target comparison that best match your biological question. Justify that choice before estimating it.
Recorded design
Researchers house 64 calves in eight pens, with eight calves per pen. In each pen they randomly assign four calves to each milk-feeding strategy. An identification-controlled feeder gives each calf its assigned strategy, so pen-mates can receive different strategies. Researchers measure baseline weight before assigning strategies, then weigh each calf at weeks 2, 4, 6, 8 and 10. Each calf stays on the same strategy. Calves share their pen environment, and the five weights from one calf describe that same calf over time.
Researchers assign feeding strategy to each calf: 64 calf assignments, 32 per strategy. They do not assign one strategy to an entire pen. The eight shared pens and repeated calf weights still need to be considered when describing uncertainty.
Use data-dictionary.qmd for column meanings and provenance.md for source information. You do not need private simulation parameters to choose a question.
Information to collect
Seek biological context for milk allocation and weaning, how individual electronic feeder settings work, and what kind of growth difference would be useful. Check which aspects of the source experiment are verified and which belong only to the invented teaching design. Use a small number of useful sources and distinguish verified information from provisional assumptions. An inaccessible paper is not a reason to miss a milestone: use the factual case metadata and seek an accessible source or instructor clarification.
Your analytical responsibilities
- Choose one biological question. State what you will measure or calculate, which groups or conditions you will compare, the units and the time period.
- Draw what researchers did: how animals or groups received treatments, or how groups were selected for observation. Show which rows describe the same animal, pen or pasture group. Explain how your analysis will allow for those shared measurements.
- Complete
analysis-plan.qmdand submit it in teaching Week 10 before fitting. Choose a model for the possible outcome values and explain every term. Record your reasons, checks and later changes. - Write your own fitting and estimation code after the relevant methods have been taught. Class examples and package documentation can help with syntax, but your group must explain the choices for this study.
- Compare one scientifically plausible alternative. For example, ask whether a different account of shared measurements or change over time alters the answer to the same question. Explain why you chose that comparison.
- Show the estimate, an interval describing its uncertainty and one clearly labelled figure. Say what the estimate means for the animals or groups studied, and what remains uncertain. Label all supplied observations as synthetic.
- Propose a feasible new experiment. Say whether researchers would assign treatments to animals, pens or pasture groups, what they would measure, and what difference would matter. Compare feasible study sizes under two sets of assumptions using
prospective-planning.md.
Keep the scope manageable. Ask for help with the design or software, while explaining your own decisions. If you saw earlier complete case outputs, record that and agree a new analytical question with the instructor.
Milestones
| Date | Task |
|---|---|
| 7 October | Choose a provisional case and question. |
| 12–18 October, calendar week 42 | Revision and one-page scoping brief, capped at 2½ hours per student; no fitting. |
| 21 October | Question/design feedback and feasible information route. |
| 28 October | Audit recorded units, allocation/sampling and metadata. |
| 4 November | Confirm question and dataset; supplied data remain a fallback if another route fails. |
| 11 November, teaching Week 10 | Submit your analysis plan before fitting, then revise after feedback. |
| 18 November morning, Unit 11 | Fit your justified analysis and gather checks/alternative evidence. |
| 18 November afternoon and 25 November | Refine estimates, communication, follow-up design and reproducibility. |
| 2 December, Unit 14 | Present and discuss. |
Suggested educational outputs: a one-page scoping brief, analysis plan, 3–5-page reproducible report including the main figure/follow-up design, and an 8-minute presentation plus 5 minutes of questions. Final assessment requirements and deadlines will be confirmed separately.
Working files and support
Open the group-projects workspace supplied in class. This case folder contains its unchanged CSV, dictionary, provenance, import/inspection starter, planning prompts and templates. Set the R working directory to the case folder and run starter.R; it uses readr and dplyr and prints the observed table/column summaries only. Create your own analysis script.
If computing is unavailable, after the analysis plan is submitted the instructor can supply outputs for your declared analysis. Acknowledge that support and interpret/check the outputs yourself. Generic class examples and instructor troubleshooting provide proportionate technical help without requiring unsupported model complexity. Retain a working copy of your analysis plan and record the sources/help informing your decisions.